The study enrolled subjects with low to high coronary artery disease based on their Gensini score (calculated by quantitative coronary angiography), aged between 42 and 78 years, from the Coronary Assessment in Virginia (CAVA) cohort. In this study, older age was defined as ≥65 years, consistent with commonly used thresholds in clinical and epidemiological research (24) https://www.nih.gov/nih-style-guide/age. Participants were recruited through the Cardiac Catheterization Laboratory at the University of Virginia Health System in Charlottesville, VA, USA, following physician referrals for medically necessary cardiac angiography. Written informed consent was obtained from all participants before enrollment, and the study received approval from the UVA Human Institutional Review Board (IRB No. 15328). Peripheral blood mononuclear cells were collected from all participants as described in our earlier studies (9, 10, 21), and clinical features, including CAD severity (Gensini score), sex, age, statin treatment, and diabetes status, were documented (Supplementary Tables 1A–C). PBMC Samples that failed to meet quality control criteria, including cell viability, were excluded from the analysis (n=4), and the final study analysis included a total of 61 samples.

The methods for performing Quantitative Coronary Angiography (QCA) and the calculation of the Gensini score have been outlined in detail in studies conducted by our group (9, 10, 21). Subjects with a Gensini score ≥32 were classified as high CAD, and subjects with a Gensini score ≤6 were classified as low CAD. Comprehensive analyses of Gensini scores highlighting statistically significant differences by CAD severity (Low CAD, High CAD) and age group (Younger and Older), together with associated demographic and clinical characteristics, are presented in Supplementary Tables 1A through 1C.

Peripheral blood was collected into BD K2 EDTA tubes and processed at room temperature within 60 min, as described previously (5, 9). PBMCs were isolated by Ficoll-Paque PLUS density gradient centrifugation using SepMate-50mL tubes (#85460, StemCell Technologies) following the manufacturer’s instructions. Cells were counted and cryopreserved in 90% FBS + 10% DMSO. To minimize batch effects, we processed at least eight samples in parallel, thawed them at 37°C, centrifuged them at 400×g for 5 min, and resuspended them in cold staining buffer. Viability and counts were measured using the BD Rhapsody Scanner. The scRNA sequencing was performed using Antibody-seq (Ab-seq) on the BD Rhapsody high-throughput single-cell analysis platform, in which antibody-derived tags (ADT) and single-cell gene expression libraries are simultaneously captured from the same individual cells. We used a customized 51 antibody, 487 immune gene panel, including molecular barcoding, an approach we have previously published and validated in multiple studies (9, 10, 21, 25).

Of 65 samples, 61 met QC criteria (viability > 80%). Each sample was barcoded with the BD Sample Multiplexing Kit, washed, pooled, stained with a 49 validated antibody surface marker panel and loaded onto Rhapsody nanowell plates using four samples per plate. Library preparation was performed as previously described (9, 10). Briefly, primed plates were loaded with cells at 800–1,000 cells/μL and incubated on a thermomixer (37°C, 1,200 rpm, 20 min) for reverse transcription. Exonuclease I treatment was followed (37°C, 30 min), after which the plate was heated to 80°C for 20 min. cDNA libraries were prepared according to the BD protocol described by Vallejo et al. (16). Library quality and quantity were assessed using the TapeStation and Qubit systems. Sequencing generated ~60,600 reads per cell on the Illumina NovaSeq platform, with pooling and depth optimized for S1 and S2 100-cycle kits. FASTA/FASTQ files were processed through the Seven Bridges Genomics pipeline. RNA and ADT assays were normalized using the LogNormalize (scale factor 10,000) and centered log-ratio (CLR) methods, respectively. Doublets were filtered with the DoubletFinder package (26).

Analyses were performed in R v4.1.0 using Seurat v4 (27). All 49 ADT features and the top 200 variable RNA features identified using the “vst” method were scaled and subjected to PCA. The first 20 of 50 components were used for batch correction with Harmony v0.1.1 (28). Batch-corrected embeddings were integrated using FindMultiModalNeighbors (WNN), followed by Louvain clustering at a resolution of 1, and t-distributed stochastic neighbor embedding (t-SNE) made with ADT expression was used to visualize the clusters. Clusters with fewer than 100 cells were excluded. For myeloid cells, non-expressed markers were removed, and both modalities were normalized and analyzed by PCA, Harmony, FindMultiModalNeighbors, and FindClusters (resolution=1.7). The clusters were annotated into three major and 13 smaller subsets of monocytes.

Spearman’s rank correlation was performed using the “cor.test” function from the stats package in R (29). This analysis was used to assess associations between age, clinical variables, and chemokine receptor expression across monocyte subsets, as well as relationships between monocyte cluster proportions and either age or Gensini score. For each patient average expression levels of chemokine receptors (CXCR3, CXCR4, CXCR5, CCR2, CCR4, CCR5, CCR6, and CCR7) were calculated separately within the specified monocyte subsets. Correlation analysis results and expression patterns were visualized using either radar plots, heatmaps and/or bubble plots.

Matched PBMC samples used in the previous Ab-seq study were obtained for validation experiments (n=7 high CAD and n=5 low CAD). The detailed staining protocol was described previously (9). Briefly, Samples were collected from older individuals with high and low CAD. Thawed PBMCs were washed with 10 mL of 1× PBS and centrifuged at 400 g for 10 minutes at room temperature. Cell viability and counts were measured using a hemocytometer and a Cellometer (Nexcelom). At least 1.5 × 10⁶ cells per well were incubated with Human TruStain FcX™ (BioLegend) for 10 minutes at 4°C, followed by Live/Dead Blue staining (Invitrogen) for 30 minutes at 4°C. Cells were then washed and stained with a surface antibody panel for 40 minutes at 4°C. Anti-CD3, Anti-CD19, and Anti-CD56 antibodies were used to exclude T, B, and NK cells. Monocyte subsets were identified as CD14⁺CD16⁻ (cMo), CD14⁺CD16⁺ (iMo), and CD14⁻CD16⁺ (nMo). Data were collected on a Cytek Aurora spectral flow cytometer and analyzed with FlowJo (BD Biosciences). Fluorescence-minus-one (FMO) controls from healthy donors were used to set gating.

Differentially expressed genes (DEGs) were identified using the FindMarkers function in Seurat (test.use = “MAST”, logfc.threshold = 0.25 or 0.1, min.pct = 0.25 or 0.1). The logfc.threshold and min.pct values used for each analysis are mentioned in the legends. Genes with adjusted p< 0.05 were considered significant. Pathway analysis was performed with the clusterProfiler package (30) using Gene Ontology Biological Process terms, and separately for up- and down-regulated DEGs.

A generalized linear mixed model (GLMM) was used to evaluate differences in myeloid cluster proportions, while accounting for both fixed and random sources of variability, using the lme4 package v1.1-31 (31). This modeling framework allows group-level comparisons while accounting for variability across individuals and adjusting for patient-level factors that may differ between groups. Patient, sex, diabetes, and statin use were considered as random effects when comparing Older_LowCAD versus Younger_LowCAD, Younger_highCAD versus Younger_LowCAD and Older_HighCAD versus Older_LowCAD. We analyzed the cMo_CCR6 validation data using the Mann-Whitney U test in Prism (GraphPad Software, Inc. version 10.1.0), data shown as mean ± SE (p-value < 0.05 was considered statistically significant).

To visualize the relationship between marker expression and Gensini score, we used smooth plots. To assess whether this association differed by age, we fitted generalized additive models (GAMs) using ‘mgcv’ R package (32) with either age-specific or shared smooth terms for Gensini score, while accounting for baseline differences between age groups. Models were compared using ANOVA (F-tests), and the resulting p-values were used to assess age-dependent effects.

We collected peripheral blood mononuclear cells (PBMC) samples from 61 subjects aged between 42 and 78 years who were recruited in the Coronary Assessment in Virginia (CAVA) cohort (9). We performed single-cell high-dimensional antibody sequencing (Ab-Seq) on PBMCs from each subject using the BD Rhapsody platform to comprehensively study changes in monocyte-specific chemokine receptor markers across age and coronary artery disease (CAD). (Figure 1A). Subjects were stratified by age, older (≥65 years, n=33) and younger (≤64 years, n=28), and by CAD severity using the Gensini score, where a GS ≤6 =low CAD and a GS ≥32 =high CAD (Figure 1B). Covariates included sex, diabetes status, smoking, and statin use. Comprehensive clinical data pertaining to low CAD risk individuals (older, mean age 70.3 ± 4.0; and younger, mean age 56.7 ± 5.3, p-value < 0.0001) and older individuals with CAD group (high CAD, mean GS 62.8 ± 35.8 and low CAD, mean GS 3.3 ± 2.4, p-value < 0.0001). Detailed clinical characteristics for all age and CAD groups were shown in Supplementary Tables 1A to 1C. Our Antibody-Seq custom panel (9) included antibodies to detect five C-C chemokine receptors (CCR2, CCR4, CCR5, CCR6, and CCR7), and three C-X-C chemokine receptors (CXCR3, CXCR4, and CXCR5).

To determine whether chemokine receptor expression showed age-related patterns in monocytes, we first analyzed correlations between chemokine receptor expression levels and age across all circulating monocytes, independently of CAD status. Spearman correlation analysis of the eight chemokine receptors, visualized in a radar plot, revealed distinct age-associated trends (Figure 1C). Except for CCR5 in younger individuals (r=-0.41, p-value=0.03), all other correlations did not reach statistical significance. CCR5 and CXCR5 expression tended to increase with age, whereas expression of CXCR3, CCR2, and CCR4 was more highly correlated with younger individuals among the total monocyte population. The transcriptional profile of monocytes in older individuals (≥65 years) showed upregulation of inflammatory mediators, including IL1B, CCL5, IL32, TNF, CCL3, and CXCL8. Concomitantly, pathway enrichment analysis showed upregulation of pathways related to interleukin-1 signaling, myeloid leukocyte migration, and chemokine-mediated signaling in monocytes from older subjects (Figure 1D; Supplementary Table 1D). Collectively, these findings indicate that older age is associated with an increase in systemic pro-inflammatory genes in circulating monocytes.

Next, we asked whether individual chemokine receptors display age-dependent expression patterns within the three major monocyte populations (classical, intermediate, and non-classical). t-SNE analysis of 45,173 monocytes from all subjects revealed three distinct clusters corresponding to classical (cMo_CD14⁺, 38,953 cells), intermediate (iMo_CD14⁺CD16⁺, 2,599 cells), and non-classical (nMo_CD16⁺, 3,621 cells) subsets, showing clear separation among populations (Figure 2A; Supplementary Figure S1). In the t-SNE projection, each color denotes a distinct monocyte subset derived from classical (cMo), intermediate (iMo), or nonclassical (nMo) populations, as defined by surface expression of the canonical markers CD14 and CD16 (Figures 2A, B). Their distinct chemokine receptor protein (Figure 2B) and their corresponding gene expression profiles are shown in Supplementary Figure S4A. We observed CCR2 expression as the highest in cMo which declined gradually in iMo and nMo, whereas CXCR3 expression increased along the transition from cMo to nMo, indicating a shift in chemokine receptor distribution in cMo and nMo (Figure 2B). Together, these findings suggest coordinated, C-C and C-X-C receptor-specific expression across the three major monocyte subsets.

To investigate whether chemokine receptor expression profiles contribute to monocyte heterogeneity, we subclustered the monocytes from Figure 2A. At single-cell resolution, distinct expression patterns of CD14 and CD16 allowed us to classify each monocyte subcluster precisely. Our analysis identified 13 subclusters in total, including eight cMo, three iMo, and two nMo populations (Figure 3A). Consistent with earlier observations (Figure 2B), CCR2 expression was strongly enriched across all cMo clusters, whereas CXCR3 expression was moderately expressed in iMo and peaked in nMo, in the nMo_SLAN^hiCXCR3⁺ subset (Figures 3B, 2B). Monocyte subcluster-specific gene expression profiles (scaled average expression) were shown in Supplementary Figure S4B. To further examine how aging affects these monocyte subclusters in the absence of CAD, we focused on individuals who were classified as low CAD risk (Gensini score ≤6). We then compared the proportions of all monocyte subclusters between younger and older groups within this low-risk cohort (Supplementary Figures S2A, S2B). As shown in Figure 3C, older subjects exhibited a significant decline in cMo_CD33^hiCD163^hiCXCR4⁺ cells (FDR-adjusted p-value = 0.002) and a corresponding expansion of cMo_CD14^lowCXCR3⁺ immature classical monocytes subsets (FDR-adjusted p-value =0.007). Additionally, Spearman correlation analysis confirmed these age-associated trends, where cMo_CD14^lowCXCR3⁺ (r=0.48, p-value=0.009) and iMo_HLA-DR^intCCR2^low (r = 0.41, p-value = 0.028) subsets were positively correlated with age, whereas cMo_CD33^hiCD163^hiCXCR4⁺ cells showed a strong negative correlation (r=–0.56, p-value = 0.001) (Figure 3D). The frequencies of the remaining monocyte subclusters and their age-associated correlations were not statistically significant (Supplementary Figures S2B, S3C). Transcriptomic profiling of these age-associated subclusters revealed striking differences. Among all populations identified, only the cMo_CD33^hiCD163^hiCXCR4⁺ subset displayed significant transcriptomic differences between ages, with 68 differentially expressed genes found in older individuals (Figure 3E; Supplementary Table 1E). Additional clusters with fewer differentially expressed genes are shown in Supplementary Figure S2D. Functional pathway enrichment analysis of the subcluster indicated downregulation of immune response pathways related to leukocyte proliferation and migration, and upregulation of processes associated with mononuclear differentiation, leukocyte adhesion and regulation of ERK1 and ERK2 cascade in older low CAD individuals (Figure 3F). Collectively, these findings demonstrate age-dependent reprogramming of monocyte subclusters, characterized by a coordinated decrease in anti-inflammatory cMo_CD33^hiCD163^hiCXCR4⁺ and an expansion of CD14^low immature cMo_CD14^lowCXCR3⁺ monocytes, highlighting probable functional remodeling of circulating monocytes with increasing age.

Having established that aging alters some of the observed C–C and C–X–C chemokine receptor expression profiles on monocyte subsets, we next asked whether these changes also reflect disease progression across age. To address this, we first observed chemokine receptor expression across all monocytes and stratified participants by Gensini score. Nonlinear trend analyses using LOESS fitting revealed CAD severity–dependent variations in multiple chemokine receptor expression across total monocytes in both younger and older individuals. This analysis specifically showed an upward trend in CXCR3 (p-value=0.05) and CCR7 (p-value=0.01) expression associated with increased Gensini score in older compared with younger individuals with high CAD (Figure 4A). To better understand CAD-associated inflammatory shifts in chemokine receptor expression. We examined the relationship between age and patient-level average chemokine receptor expression in monocytes among younger (Figure 4B) and older adults (Figure 4D) stratified by low and high CAD risk. Although no associations reached statistical significance, younger individuals with high CAD exhibited positively trended correlations, specifically between CCR2 and CCR4 expression and high CAD burden (Figure 4B), suggesting early inflammatory remodeling of monocyte chemokine responsiveness. Consistent with these observations, differential gene expression and pathway analyses in younger individuals with high CAD revealed a statistically significant upregulation of TGFB1, CXCL16, IL32, CCL5, and IL15. These transcriptional changes were accompanied by enrichment of pathways related to chemokine-mediated signaling and cellular responses to chemokines, indicating heightened inflammatory activation in this group (Figure 4C; Supplementary Table 1F). In contrast, older individuals with high CAD burden exhibited downregulation of key inflammatory genes, including CXCL8, IL1B, and CCL3 (Figure 4E). Pathway enrichment analysis in this group highlighted activities associated with regulation of the endothelial barrier and negative regulation of cell adhesion molecule production, suggesting attenuation of pro-inflammatory and leukocyte–endothelial interaction pathways with age (Figure 4E; Supplementary Table 1G). Together, these findings suggest age-dependent divergence in monocyte inflammatory programming in CAD, with enhanced chemokine expression-driven inflammatory signaling in younger individuals and relative dampening of vascular wall and pro-inflammatory effector pathways in older adults.

We next assessed the frequencies of individual monocyte subclusters in both younger and older individuals stratified by their CAD status. In younger individuals with high CAD, we observed a significant reduction in cMo_CD33^hiCD163^hiCXCR4⁺ cells (FDR-adjusted p-value = 0.007), and an increase in iMo_HLA-DR^intCCR2^low populations (FDR-adjusted p-value = 0.02) (Figure 4F). In older individuals with high CAD, the frequencies of cMo_CD14^lowCXCR3⁺ (FDR-adjusted p-value = 0.0006) and cMo_HLA-DR⁺CCR6⁺ (FDR-adjusted p-value = 0.04) subsets were significantly reduced (Figure 4G). Interestingly, iMo_HLA-DR⁺CXCR3⁺ is the only population which was expanded by approximately two-fold in individuals with high CAD, an effect that was independent of age (FDR-adjusted p-value = 0.04), shown in Figures 4F, G, right panels. The remaining subclusters and their CAD-associated proportions in both younger and older individuals that were not statistically significant are presented in Supplementary Figures S3A, S3B, S3C.

To further characterize the transcriptomic profile underlying the age-independent expansion of the iMo_HLA-DR⁺CXCR3⁺ cluster in individuals with high CAD, we compared gene expression patterns between iMo_CXCR3⁺ and iMo_CXCR3^low/− cells. Differential gene expression analysis revealed significantly increased expression of the complement genes C1QA and C1QB in iMo_CXCR3⁺ cells (Figure 4H; Supplementary Table 1H). Pathway enrichment analysis indicated that CXCR3⁺ monocytes were transcriptionally enriched for pathways related to classical complement activation and leukocyte-mediated immunity. In contrast, genes downregulated in CXCR3⁺ relative to CXCR3^low/− cells were enriched for Gene Ontology terms associated with leukocyte tethering, rolling, and cell–cell adhesion, highlighting a distinct transcriptional signature associated with CXCR3 expression in intermediate monocytes (Figure 4I).

Next, we found that iMo_HLA-DR⁺CXCR3⁺ (r=0.46, p-value = 0.006) and cMo_CD33^hiCD163^hiCXCR4⁺ (r=0.36, p-value = 0.04) subsets showed robust positive associations with CAD severity, whereas cMo_CD14^lowCXCR3⁺ (r=–0.44, p-value = 0.009) and cMo_HLA-DR⁺CCR6⁺ (r=–0.50, p-value = 0.003) subsets were negatively correlated with CAD (Figure 4J). We validated using flowcytometry that there was a ~2.5-fold reduction in cMo_CCR6⁺ monocyte frequencies (p-value = 0.04) in additional older high CAD group from CAVA (n=7 high CAD, n=5 low CAD) (Figure 4K), consistent with our earlier Ab-Seq data (Figure 4G, middle panel cMo_HLA-DR⁺CCR6⁺). Overall, these results show that chemokine receptor expression profiles are varied in monocyte subsets of older adults with progressive CAD, highlighting a link between monocyte heterogeneity and disease severity.

Finally, to understand how inflammaging-related immune alterations associate with clinical parameters, we examined the relationship between chemokine receptor expression across the three major monocyte subsets and plasma markers of cardiovascular disease progression in older individuals. Using Spearman correlation analyses with systemic risk factors from the same subjects, we found positive associations of chemokine receptors with plasma LDL- and HDL-cholesterol and the inflammatory marker hsCRP (High-Sensitivity C-Reactive Protein) in older subjects within our cohort (Figure 5A). Many of the C-C chemokine receptors were positively correlated with plasma HDL-Cholesterol levels, including iMo-CCR6 (r=0.45, p-value =0.008), nMo-CCR6 (r=0.53, p-value=0.001), and nMo-CCR5 (r=0.44, p-value =0.010), and with a modest trend for cMo CCR4 (r=0.33, p-value=0.058). The C-X-C chemokine receptor CXCR5 is correlated positively with only LDL-Cholesterol in non-classical monocytes (r=0.40, p-value=0.020). Moreover, CXCR4 expression revealed a significant negative correlation with hsCRP levels for both iMo (r=0.39, p-value=0.024) and nMo (r=0.44, p-value=0.010) subsets. Additional analysis of monocyte subsets with CAD severity (Gensini score) showed distinct associations, in which CCR6 expression in intermediate monocytes positively correlated with Gensini Score (r=0.35, p-value=0.045), whereas CXCR3 expression in classical monocytes showed a strong negative correlation (r=–0.38, p-value=0.028) (Figure 5B). These findings suggest that selective remodeling of chemokine receptor expression within distinct monocyte subsets is likely linked to clinical cardiovascular risk parameters, including plasma cholesterol profiles and inflammatory markers in older individuals.